Intellectual property rights, trade agreements, and international trade

Intellectual property rights, trade agreements, and international trade

Research Policy xxx (xxxx) xxx–xxx Contents lists available at ScienceDirect Research Policy journal homepage: www.elsevier.com/locate/respol Intel...

NAN Sizes 0 Downloads 2 Views

Research Policy xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Research Policy journal homepage: www.elsevier.com/locate/respol

Intellectual property rights, trade agreements, and international trade Mercedes Campia, , Marco Dueñasb ⁎

a b

CONICET, University of Buenos Aires, Faculty of Economics, Instituto Interdisciplinario de Economía Política de Buenos Aires (IIEP – Baires), Argentina Department of Economics, International Trade and Social Policy – Universidad de Bogotá Jorge Tadeo Lozano, Colombia

ARTICLE INFO

ABSTRACT

JEL classification: O10 O34 F14

The global process of strengthening and harmonization of intellectual property rights (IPRs) systems has been intensified in the last twenty five years by the signing of trade agreements (TAs) that include chapters with intellectual property (IP) provisions and other trade-related issues. This paper provides a first exploration of whether and how the signing of TAs with IP chapters influences bilateral trade flows for a balanced panel of 110 countries and the period 1995–2013. We address methodological issues related to the assessment of the effect of TAs on bilateral trade. We use matching econometrics to evaluate the treatment of TAs with and without IP chapters. In addition, we estimate the effects of TAs on bilateral trade in a more dynamic fashion using a panel data approach based on the gravity model. Also, we perform our analysis for trade in low- and high-IP intensive products. We found that both types of TAs increase bilateral trade but TAs with no IPRs chapters have a stronger positive effect on trade. However, if we include lags to consider that TAs with IP chapters might need a longer implementation time, the net expected increase on trade is similar for both types of TAs. We also found that the effects depend on the development level of countries and on the IP intensity of products. We found a clear positive effect for developed countries, but we do not observe important gains for developing countries in all sectors and to all destinations derived from TAs with IP chapters. This raises the question of whether trade gains can compensate the effort related with IP reforms.

Keywords: Intellectual property rights Trade agreements International trade Matching econometrics Gravity model

1. Introduction The last decades have been characterized by an increasing interaction among countries, especially through trade and foreign direct investment (FDI). This was shaped since the 1990s through a process of reduction of trade and investment barriers, reinforced through the signing of bilateral, regional, and multilateral trade and investment agreements. The process of globalization is also reflected in changes in international institutional and normative aspects. The creation of the World Trade Organization (WTO) in 1994 to regulate international trade and to establish a framework for trade policies came along with several agreements demanding institutional reforms including those related with intellectual property rights (IPRs). The agreement on Trade Related Aspects of Intellectual Property Rights (TRIPS) is certainly one of the most renowned. With the signing of the TRIPS agreement, countries are bound to adopt or modify their intellectual property (IP) related legislation in

accordance to certain minimum standards. Since then, there has been a process of global strengthening and harmonization of IPRs systems (see the increasing scores of IPRs indexes in: Park, 2008; Liu and La Croix, 2015; Campi and Nuvolari, 2015), despite different countries in terms of development and capabilities might need specific types of IPRs systems (Kim et al., 2012). In addition, several countries have been tightening their IPRs systems because they have signed trade agreements (TAs) that include complex chapters covering IPRs, with IP provisions that demand higher standards of IP protection and are known as TRIPS-Plus or TRIPS+. The obligations related to IPRs are usually included along with a set of policies that need to be implemented in order to comply with the requirements of the agreement and to take advantage of its trade issues.1 This includes new areas of IPRs, such as the patenting of life forms or copyright applying to electronic content, the implementation of more extensive levels or standards of IP protection than the ones demanded by TRIPS, the adoption of new conventions not included in other TAs, or the elimination of an option or flexibility available under TRIPS

Corresponding author. E-mail addresses: [email protected] (M. Campi), [email protected] (M. Dueñas). 1 In this paper, we differentiate between “trade issues” as those that entail provisions derived from trade policies and “trade-related issues” as those that, not being directly trade issues, have a connection with them, such as provisions on investments, IPRs, services, public procurement, competition, sanitary and phytosanitary measures, dispute settlements, trade defense instruments, market access, and other dimensions, which can also affect trade relations. ⁎

https://doi.org/10.1016/j.respol.2018.09.011 Received 1 September 2017; Received in revised form 13 September 2018; Accepted 29 September 2018 0048-7333/ © 2018 Elsevier B.V. All rights reserved.

Please cite this article as: Campi, M., Research Policy, https://doi.org/10.1016/j.respol.2018.09.011

Research Policy xxx (xxxx) xxx–xxx

M. Campi, M. Dueñas

(Mercurio, 2006; Biadgleng and Maur, 2011). As a consequence, TAs are increasingly guiding the design of IPRs systems and strengthening IP protection worldwide, despite being, in principle, a trade policy. In fact, as Maskus (2015) argues, it is difficult to think that the increase in the number of TAs and the strengthening of IPRs systems are independent processes. In particular, IP-demanding countries are often developed countries (DCs), while usually developing or least developed countries (LDCs) are the ones implementing the provisions. Thus, TAs are clearly drivers of significant reform in LDCs and their implementation implies a real and complex challenge for them (Biadgleng and Maur, 2011). In this context, we ask whether and how TAs with IP chapters affect international trade. Several other question arise: are these TAs, with their effects on IPRs systems, affecting or shaping differently international trade relations compared to TAs with no IP chapters? Do they affect equally countries of different development level? Could the expected gains in trade compensate the efforts related with IP reforms for developing countries? Despite the increase in the number of TAs including IP provisions, only a few recent studies address their implications and investigate how they affect international trade. In this paper, we provide an empirical and econometric analysis of the effect of TAs with IP chapters and TAs with no IP provisions on bilateral trade flows of 110 countries, for the post-TRIPS period: 1995–2013. We use data from Kohl et al. (2016), which have information on TAs with IP provisions, and bilateral trade data from Gaulier and Zignago (2010). We first employ matching econometrics in the cross-section in order to compare the effect on two treatment groups (country pairs that signed TAs with IP chapters, and country pairs that signed TAs with no IP chapters) against a control group (country pairs that did not sign TAs). Secondly, we do a difference-in-difference analysis to estimate the impact of both types of TAs on bilateral trade flows, employing individual level panel data analysis with fixed effects and a gravity model framework. We analyze the effect on bilateral trade of manufactures of high- and low-IP intensity, and we control whether the effect is heterogeneous for countries of different development level. Additionally, we present several robustness checks. Overall, we found that both types of TAs increase bilateral trade. However, TAs with no IP chapters have a stronger positive effect, while the effect from TAs with IP chapters increases if we include lags to consider that they might need a longer implementation time. The net expected increase on bilateral trade is similar for both types of TAs. Also, we found that the effects of both types of TAs depend on the development level of the signatory countries and on the IP intensity of products. Both types of TAs increase trade flows from developed countries to both developed and developing countries, and the increase derived from TAs with IP chapters is higher. Instead, trade flows from developing countries are mostly increased by the signing of TAs with no IP chapters. While we found a clear positive effect for DCs, we do not observe significant gains for LDCs in all sectors and to all destinations. This raises the question of whether the relatively small trade gains derived from TAs with IP provisions can compensate the effort related with IP reforms for LDCs. The paper is organized as follows. Section 2 provides the motivation, reviews the literature, and discusses several methodological challenges related with the estimation of the effect of TAs on trade. Section 3 analyzes the evolution and diffusion of TAs. Section 4 presents the econometric estimations. Finally, Section 5 concludes.

Baier and Bergstrand, 2007; Medvedev, 2012). Theoretically, given that TAs remove domestic barriers to trade, most authors expect a positive effect on trade flows of signatory countries. However, TAs can lead to trade diversion rather than to trade creation, and can be a substitute for the full implementation of WTO rules, which cannot necessarily lead to a positive effect on trade (Kohl et al., 2016). Thus, from a theoretical perspective, the effect of TAs on trade flows is ambiguous. Likewise, the findings of empirical approaches are rather mixed. On one of the first studies, Rose (2004) found little evidence that the General Agreement on Tariffs and Trade (GATT) and the WTO positively affect trade. Later, other authors showed that the WTO and other TAs had a positive but uneven impact on trade (see Subramanian and Wei, 2007; Cheong et al., 2015, for the case of the WTO). A reason behind these mixed empirical findings is the existence of several methodological issues that make the estimated effect of TAs on trade flows highly sensitive to the specification of the model and the groups of countries or years chosen (see, Magee, 2003, for a discussion). A major methodological issue and a plausible explanation for these non-concluding results derives from the endogeneity of TAs. There exist clear endogenous reasons for countries to engage in TAs, which are also likely to be correlated with the levels of trade and country characteristics (see: Magee, 2003; Baier and Bergstrand, 2004, for studies on the determinants of TAs). Due to the presence of endogeneity, the effect of TAs is usually under- or over-estimated. In particular, endogeneity is acknowledged as an important problem when dummy variables are used to estimate the effects of TAs using gravity models with crosssectional data and ordinary least squares (OLS) estimation method. A few contributions address the issue of endogeneity using gravity frameworks estimated with fixed effects, finding that previous analysis underestimated the impact of TAs. Baier and Bergstrand (2007) showed that most cross-section gravity estimations of the effect of TAs on bilateral trade flows result in biased, unstable, and underestimated effects, and that the major source of endogeneity derives from omitted (selection) bias. They argue that better estimations are obtained with gravity equations using panel data with bilateral fixed and country and time effects or differenced panel data with country and time effects. In a more recent article, Baier and Bergstrand (2009) used matching econometrics to estimate the long-run effects of TAs on bilateral trade flows, avoiding the bias introduced by non-random selection and nonlinearity. They conclude that estimations are more stable across years and have more plausible values than cross-section estimates of gravity equations with OLS. Building on this contribution, Falvey and Foster-McGregor (2017) used matching econometrics to investigate the impact of preferential trade agreements (PTAs) on trade flows. They considered two sequential decisions of PTAs: first, whether two trading partners should form a PTA and, second, if they do, how broad that agreement should be. They found that distance, common language, common boarder, and GDP are significant for both decisions, but often have opposing effects on each one. They estimated a dose response function that relates the trade change due to PTA treatment to the breath of the PTA adopted, and they showed that it exhibits an inverted u-shape. Notice that Baier and Bergstrand (2007, 2009) and Falvey and Foster-McGregor (2017) estimate the log of the sum of the bilateral trade flows between partners. Using this dependent variable, they are able to observe the average effect of TAs on country pairs but they cannot address possible uneven effects on trade partners. Cheong et al. (2015) showed that the more similar the partner countries are in terms of size, income, or location, the larger the increase in intra-bloc trade is under a TA. Also, they showed that the gain for LDCs from a TA among themselves is about two and a half times that from partnering with DCs. Likewise, there might exist differences derived from the level of technology and development of trading countries. Shin et al. (2016) showed that IPRs may act as an export barrier to trade, discouraging exports from LDCs that are in the process of catching-up in terms of their levels

2. Motivation and methodological aspects The remarkably increase of different types of TAs during the last twenty five years has spurred the interest of economists. A large literature analyzes the effect of TAs on international trade, FDI, economic integration, and economic growth (see, for example, Krugman, 1993; 2

Research Policy xxx (xxxx) xxx–xxx

M. Campi, M. Dueñas

of technology. They argue that while recent IPRs reforms have facilitated global trade, they have not helped promoting exports of LDCs to countries that strengthen their IPRs systems. Thus, IP protection creates a distributional bias in favor of exporters from DCs relative to those from LDCs. Considering this evidence, we allow for the existence of heterogeneous effects on trade partners. Increasingly, TAs include not only trade issues but also provisions on investments, IPRs, services, public procurement, competition, sanitary and phytosanitary measures, dispute settlements, trade defense instruments, market access, and other dimensions, which can also affect trade relations. Despite TAs are different in scope, content, and design, in general, the literature considers simply the number of TAs in force. Addressing this issue, several authors started building databases in which TAs are classified by their content. Dür et al. (2014) created a database that considers the differences in the design of PTAs and they analyzed whether and to what an extent PTAs affect trade flows. They find that, on average, PTAs increase trade but the effect depends on the depth of the agreements. Likewise, using a gravity model, Kohl et al. (2016) studied how heterogeneous 296 TAs signed between 1948 and 2011 stimulated international trade. They found that their heterogeneity is relevant for explaining trade flows and that the degree to which countries negotiate comprehensive TAs depends positively on their development level and on the number of WTO members on the agreement. Similarly, Kohl and Trojanowska (2015) addressed the heterogeneous nature of TAs using matching econometrics to evaluate their impact. They found that the magnitude and significance of the treatment effects positively depend on the extension the agreements. Similarly, Hofmann et al. (2018) created a database of PTAs that provides a detailed assessment of their content, showing that the depth is positively correlated with the intensity of trade flows. Using this data, Mattoo et al. (2017) evaluated the impact of the depth of PTAs on trade, controlling the effect of other determinants of trade flows, and assessing endogeneity problems with a gravity model. They found that deep agreements lead to more trade creation and less trade diversion than shallow agreements.

(EU), or the Europe Free Trade Association (EFTA), and have chapters on IP-related issues have significant impacts on trade flows of the members, especially for middle-income countries, but also for high- and low-income countries in some particular sectors. Our analysis complements the results of Maskus and Ridley (2016) because we estimate the effects of TAs with IP chapters on bilateral trade flows considering the interaction with the development level of signatory countries and comparing them with the effect of TAs with no IP chapters. In general, IP chapters include access to multilateral conventions on the protection of IPRs, such as the Rome Convention, the Paris Convention, the Bern Convention, the WIPO Copyright Treaty, the WIPO Phonograms Treaty, and the 1991 Act of the UPOV Convention. Also, they might have references to the scope of IPRs, including specific provisions in relation to substantive standards of protection, pharmaceuticals, geographical indications, the patenting of life forms, copyrights applying to electronic content, establishment of minimum duration of IPRs – for example, the author's lifetime plus 70 years for copyrights that extends the minimum of plus 50 years of the TRIPS – the elimination of options or flexibility available under TRIPS, and enforcement, including criminal and civil penalties for infringement. Most of these provisions exceed the minimum standards of the TRIPS and this is why they are known as TRIPS-Plus or TRIPS+. In most cases, IP provisions are legally enforceable as they imply a clearly defined obligation and this effectively bind the parties (Horn et al., 2010). Certainly, the content of IP chapters of different agreements is not homogeneous. For example, the US, the EU, and EFTA, which are the main drivers of this type of TAs, show differences. While the United States emphasizes on patent and copyright protection, the EU and EFTA emphasize even more the protection of geographical indications (Maskus and Ridley, 2016). US treaties usually include requirements for test data confidentiality for pharmaceuticals and chemicals, while only more recently the EU began demanding similar rules (Biadgleng and Maur, 2011). In addition, TAs with the EU contain, mostly, the general principle of adhering to the highest standards of protection and ratification of WIPO treaties and UPOV 1991. This has implications for many LDCs that have sui generis systems in place designed according to UPOV 1978, which already comply with the minimum standards demanded by the TRIPS. Thus, LDCs involved in TAs with IP chapters – in particular with the US, the EU, or EFTA – have more stringent IPRs systems compared to LDCs that did not sign this type of TAs. This sheds light on another relevant issue related to the endogeneity of IP provisions of TAs. As argued by Maskus and Ridley (2016), IP provisions can be for most LDCs second-order negotiating concessions that they would not ordinarily select as a matter of endogenous policy. Similarly, most LDCs were reluctant to adopt several measures of the TRIPS and their adoption was not an endogenous response to domestic innovation (Delgado et al., 2013). Therefore, both IP provision in TAs and IP reforms derived from TRIPS can be considered exogenous, up to a certain extent, at least for LDCs. Given that most TAs with IP provisions involve an IP-demanding country – usually a DC – and one or several countries implementing the IP reforms – often LDCs – not considering the differences in the development levels of countries can provide an incomplete or biased assessment of the effect of TAs with IP chapters. At least two other methodological issues arise when evaluating the impact of TAs with IP chapters on trade. One of them is related with the difficulty of separating the effect of IP chapters from the effect of trade issues, both included in the same agreement. The other one is related with the implementation of IP provisions, which has been rarely studied. Most IP provisions are discussed and defined during the negotiation of the agreement, before the date of entry into force. Maskus and Ridley (2016) argue that most IPRs chapters require specific compliance dates, upon or soon after the date of a treaty's entry into force. In this context, they argue, the binary nature of this policy variable is appropriate. Nevertheless, Biadgleng and Maur (2011) claim that IP demands

2.1. Evaluating the effect of IP provisions Until recently, the effect of extra-trade provisions and, in particular, those related with IP protection, had been marginally studied. IP provisions can have an effect on trade flows beyond trade issues because changes in IPRs systems may influence returns to innovation, affecting decisions of firms to trade in different markets. Schneider (2005) showed that strong domestic IPRs might hurt innovation in LDCs because most of their innovations are imitations or adaptive. Also, stronger IPRs and enforcement in DCs might curb imports from LDCs, as the latter's exports would be excluded if they are found to be too imitative and infringing (Shin et al., 2016). Theoretically, Maskus and Penubarti (1995) argue that stronger IPRs systems are expected to have contrary effects on trade. On the one side, firms should be encouraged to export patentable goods to countries with stronger IPRs because the risk of imitation decreases. Simultaneously, stronger IPRs increase the market power of firms, which may encourage them to behave in a monopolistic way, increasing prices and reducing sales. The net result will depend on the sectors and the level of development of trading partner countries. Similarly, different models have concluded that the effect of IPRs on trade is ambiguous and that whether the strength of IPRs is desirable, cannot be answered theoretically (Grossman and Helpman, 1990; Helpman, 1993; Grossman and Lai, 2004). Therefore, empirical analysis are needed to disentangle the effect of stronger IPRs on bilateral trade flows. To our knowledge, the first empirical contribution investigating the effect of IP-related provisions in TAs is the one developed by Maskus and Ridley (2016) who study the role of PTAs with complex chapters on IPRs on the magnitude and composition of trade. They found that PTAs in which one partner is the United States (US), the European Union 3

Research Policy xxx (xxxx) xxx–xxx

M. Campi, M. Dueñas

Fig. 1. Evolution of the number of signed trade agreements, 1948–2011.

Fig. 2. Network of countries with trade agreements with no IP chapters (in blue) and with IP chapters (in red). 1995 (left) and 2010 (right). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)

increasing since 1992.3 This process is expected to have several implications for signatory countries but also for other countries that have trade relations with them. The TRIPS agreement extends the principles of “national treatment” and of “most-favored-nation” to IPRs for all WTO members. This means that WTO members must be treated equally to the residents regarding IP protection and that IP reforms – including those that derive from a TA – are extended to all other countries connected with the members of the TA once they are implemented. Certainly, this reinforces the process of harmonization and strengthening of IPRs systems. In particular, TAs are inducing the diffusion of stronger IPRs systems to LDCs because TAs with IP chapters usually involve a DC or developed area and one or several LDCs. Fig. 2 shows the increase between 1995 and 2010 of bilateral connections through TAs with no IP chapters (in blue) and TAs with IP chapters (in red). Most countries with TAs with IP chapters are linked with the EU or the US. In addition, several developing countries, such as Chile or South Africa, have signed mainly TAs with IP chapters and most of them with the EU or the US as a partner. From a different perspective, Fig. 3 illustrates how countries have become more integrated through different types of TAs. We observe that TAs have been increasing over time with an expansion of members as a consequence. While the network was disconnected in 1995, it looks highly connected in 2010, although it still has strong regional features.

often require a relatively extensive review of legislation, regulations, and practices, which creates a challenge for econometric studies that use information contained in TAs. The authors found that TAs are clearly drivers of significant reform in countries and that the implementation challenge for LDCs is real and complex. In the econometric analysis, we adopt different strategies to deal with endogeneity, selection bias, and other methodological issues. To separate the effect of trade issues and IP provisions, we use matching econometrics and compare the effects of TAs with and without IPRs chapters. For further control of endogeneity, we perform a differencein-difference analysis employing individual level panel data analysis with fixed effects and a gravity model framework. To consider the implementation time of IP provisions, we include lags of TAs. We use interaction variables between TAs and the level of development of countries to consider possible heterogeneous effects. In addition, we control for the level of IP intensity of products. Finally, we carry out several robustness checks. 3. The diffusion of trade agreements Countries around the world have been increasingly signing different types of TAs.2 Fig. 1 shows the evolution of the number of TAs between 1948 and 2011. We observe that only a few TAs were signed between 1948 and 1991, and that their diffusion is a phenomenon that clearly characterizes the last twenty five years. Moreover, the number of TAs with IP provisions, mostly legally enforceable, has been steadily

3 Our data distinguishes between provisions that can and cannot be considered to be legally enforceable commitments in a court of international law. Kohl et al. (2016) classifies TAs including IPRs provisions as legally enforceable only if the undertaking specified at least some obligation that is clearly defined, and that is likely to effectively bind the parties, as defined by Horn et al. (2010). It may also be the case that undertakings are not legally enforceable because they are explicitly excluded from the trade agreement's dispute settlement procedures.

2

This includes reciprocal trade agreements between two or more partners – free trade agreements and customs unions – and PTAs – unilateral trade preferences that comprise Generalized System of Preferences schemes (under which DCs grant preferential tariffs to imports from LDCs), as well as other nonreciprocal preferential schemes granted a waiver by the General Council. See: https://www.wto.org/english/tratop_e/region_e/rta_pta_e.htm, accessed on January 2017. 4

Research Policy xxx (xxxx) xxx–xxx

M. Campi, M. Dueñas

Fig. 3. Trade agreements without IP chapters (in blue) and with complex IPRs chapters (in red). Evolution of the network of signatory countries. See Appendix A for the list of countries. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)

The networks show the existence of communities or clusters of countries, which usually are trade areas or geographical neighbors. For all the years, there are clearly distinguishable clusters of countries linked through TAs with no IP chapters (in blue). Some of these clusters include LDCs and are not linked to DCs. Instead, most clusters of countries with TAs with IP chapters include DCs. In 1995, we observe three clusters of countries with TAs with IP chapters composed by the EU, a group of Eastern European countries, and a third smaller cluster composed by the US, Canada, and Mexico. In 2000, a fourth cluster composed of a group of African countries appears and the other three clusters get more dense. In 2005, the network grows and several Asian countries sign TAs with IP chapters. Finally, in 2010, we observe a more connected network, more dense clusters, and a new cluster composed of a group of Asian countries.

corresponding chapter in order to consider heterogeneity among price variations for different sub-sectors.4 We use data from Kohl et al. (2016) that contains TAs signed between 1948 and 2011, and distinguish TAs with no IP chapters and with both legally and not legally enforceable IP chapters. We use TAs with legally enforceable IP chapters in our variable indicating the presence of IP provisions as we are interested in TAs that entail a compulsory implementation of IP reforms. Although not legally enforceable IP provisions might also lead to IP reforms, we have no certainty of this. Conversely, legally enforceable IP provisions bind countries to implement them in order to comply with the obligations of the agreement. We have modified the database to consider that the accession agreements enacted to become part of the European Union include legally enforceable IP provisions.5 We consider TAs signed before 1995 that are

4. Econometric estimations 4 The deflators are available at the US Bureau of Labor Statistics, http:// www.bls.gov/web/ximpim/beaexp.htm, accessed on January 2017. 5 We have analyzed legal documents of the European Union to determine this. See EUR-Lex: http://eur-lex.europa.eu.

We use bilateral trade flows from Gaulier and Zignago (2010) for the years 1995–2013 and a balanced panel of 110 countries. We deflate the data by the US imports price index applying the index of each 5

Research Policy xxx (xxxx) xxx–xxx

M. Campi, M. Dueñas

these two different treatments (TAs with and without IP chapters). We follow Abadie and Imbens (2006) who developed a matching methodology that stresses on how to properly define a control group of countries that can be compared and paired with treated countries. To do this, the matching mechanism simulates random assignments based on a set of characteristics (x) of the country pairs in both groups that might be as much similar as possible. Therefore, it allows to derive the change in the expected value of total bilateral trade taking as a reference the non-treated and the treated group of countries. Accordingly, the first case is called the average treatment effect (ATE), while the second one is called the average treatment effect of the treated (ATET). These two measures are relevant for the treatment evaluation of TAs. The ATE is relevant when the treatment has universal applicability, while the ATET is relevant to consider the counterfactual, i.e. the average gain from the treatment for the treated (Cameron and Trivedi, 2005). From the analysis above, it is clear that TAs have been rapidly diffusing in the world trade network. The evaluation of this phenomenon is not completely captured by the ATET, which instead allows to assess the impact of TAs on the treated. In contrast, with the ATE we can study the effect at the population level in order to take into account the potential outcome for country pairs that still lack TAs. We derive the ATE and the ATET for the group of country pairs with TAnip and the group of country pairs with TAip. The treatment effects for the treatment groups are defined as:

Table 1 Number of links, bilateral trade flows, and differences for selected years. Countries with: TAs without IP chapters (TAnip), TAs with IP chapters (TAip), and no TAs.

nip

1995

2000

2005

2010

Number of links

TA TAip No TAs

525 515 10,950

671 823 10,496

740 1286 9964

850 1648 9492

Average of the ln of trade

TAnip TAip No TAs

10.89 12.52 8.58

10.79 12.31 8.50

11.17 11.94 8.78

11.57 11.53 9.01

Difference in ln between

TAnip and No TAs TAip and No TAs

2.31 3.94

2.29 3.81

2.39 3.16

2.56 2.52

still in force in our time period, and we extrapolate the list of TAs in force for the years 2012 and 2013. In order to evaluate the economic effect of TAs, we follow two complementary strategies. Firstly, we employ matching econometrics in the cross-section and, secondly, we use a difference-in-difference technique applied to the whole time period and a gravity framework. In both cases we define the control group as all those country pairs without TAs, and two different treatment groups: one includes countries that have signed TAs without IP chapters, and the other one includes countries that have signed TAs with IP chapters. We use two treatment groups because we want to compare the effect of the two types of TAs using a common control group. Therefore, we define two different policy variables: one for TAs without IP chapters TAijtnip , and another one for TAs with IP chapters TAijtip . These are dummy variables that take the value of one from the entry into force of a specific TA between the country pair ij. Because TAs have different dates of entry into force and are signed by different country pairs ij, both control and treatment groups change over time. Table 1 shows several statistics for the three groups of countries and for selected years. We observe that the number of bilateral trade links (strictly positive flows) with any type of TAs increases over time, while the number of bilateral trade links of countries with no TAs decreases. This reflects the increase in the number of signatory countries. On average, country pairs with any type of TAs trade more than country pairs with no TAs. The difference between the control group and the two treatment groups changes over time. In the case of TAnip, the difference increases, while in the case of TAip, the difference decreases. Interestingly, a possible explanation for the lower average trade levels for country pairs with TAip can derived from the fact that the number of links increased by 3.2 times between 1995 and 2010. Instead, the number of links of countries with TAnip increased by 1.4 times. We perform the estimations for total bilateral trade and trade of products of different IP intensity. We use the classification of Delgado et al. (2013), which divides products from the Standard International Trade Classification (SITC), Revision 3, into two categories: high-IP and low-IP intensive products. Using this information, we classify products in the Harmonized System Codes 1992.6 Appendix B presents the detailed list of products of different IP intensity.

ATEnip (x ) = E [w˜ 1nip

w˜ 0 |X = x ];

ATET nip (x ) = E [w˜ 1nip ATEip (x ) = E [w˜ 1ip

ATET ip (x ) = E [w˜ 1ip

w˜ 0 |TAnip = 1, X = x ];

w˜ 0 |X = x ];

w˜ 0 |TAip = 1, X = x ];

(1) (2) (3) (4)

where w˜ is the natural logarithm of total trade between country pairs ij in a given year; and w˜ nip refers to total trade of countries that have signed TAnip, w˜ ip is total trade of countries that have signed TAip, w0 is total trade of countries that have not signed TAs; and, X is a random vector of dimension k of continuous covariates distributed on k , with compact and convex support x. For the matching mechanism, we employ a logit model to find the propensity scores in which the independent variable is whether country pairs have (or do not have) a TA (with and without IP chapters), using as covariates countries’ GDPs, geographical distance, contiguity, common language, and a set of dummies characterizing the development level of trading partners. More formally, the set of economic characteristics x is a vector of k = 7 dimensions that we define as x = { ln(GDPi · GDPj), ln(d), contig, comlang, Gdc↔dc, Gdc↔ldc, Gldc↔ldc}. Once propensity scores are determined, we use three nearest neighbors in the control group. See Appendix C for a complete description of the variables and sources. Baier and Bergstrand (2009) extensively discuss three conventional assumptions of the matching methodology for the treatment evaluation of TAs. The ignorability assumption is that, conditional to x, TAs are independent of the outcomes w˜ 1 and w˜ 0 . The overlap assumption claims that the distribution of x for the treated and the untreated trade partners have a common support. And, the assumption of stable-unit-treatment-value that implies that conditional to x: i) the treatment (TAs) is the same for all treated countries, i.e. there are no multiple versions of TAs, and ii) there are no network effects (no interference) such that a TA between a country pair ij influences the outcome of another country pair without a TA. These assumptions are difficult to meet, in particular, the third one. To deal with this, Baier and Bergstrand (2009) use the strategy of redefining the set of covariates x. Their approach departs from the theoretical hypotheses of general equilibrium, which borrows from the microfounded model proposed by Anderson (1979). However, the

4.1. Matching estimations We use a matching approach in order to have a good assessment of the treatment effect of TAs in the cross-section. Another motivation derives from the difficulty in evaluating the impact of IP provisions and of trade issues because they are included in the same TA. Thus, by creating two treatment groups and comparing them with a common control group, we are able to compare in relative terms the effects of 6 Conversion tables can be found at: https://unstats.un.org/unsd/trade/ conversions, accessed on January 2017.

6

Research Policy xxx (xxxx) xxx–xxx

M. Campi, M. Dueñas

Fig. 4. Estimated average treatment effect (ATE, upper panel) and estimated average treatment effect on the treated (ATET, lower panel) using matching econometrics for bilateral trade flows (in ln): total, high-IP intensive products, and low-IP intensive products. Shaded areas correspond to the 95% confidence intervals.

network effects in the international trade system are too strong to be reconciled with these micro-assumptions (see the role of network structure in Arpino et al., 2017). In fact, a relevant number of studies show that the international trade network has strong signals of complexity, which implies that departing from the micro to explain the macro phenomena is very difficult (Serrano and Boguñá, 2003; Dueñas and Fagiolo, 2013; Almog et al., 2017). Therefore, we prefer not to redefine the covariates (x) because in doing so one would accept that any trading country pair may be equally affected by any other trading country pair, which ignores the topology that governs the international trade network. Of course, we are aware of the caveats at using the matching methodology in order to do the treatment evaluation. Fig. 4 shows the estimation results. In the upper panel, we show the estimated ATE for total bilateral trade (left), trade of high-IP intensive products (middle), and trade of low-IP intensive products (right). In all cases, we observe that the average treatment effect is significant and for both treated groups (country pairs with TAnip or country pairs with TAip) estimates higher average trade levels than for the control group (country pairs with no TAs). We observe no statistically significant differences between the estimated average bilateral trade of the two treatment groups with respect to the average trade of the control group. The lower panel of Fig. 4 shows the estimation results of the ATET. Likewise, the estimation results show that both types of TAs actually increase the levels of total trade and of products of high- and low-IP intensity for both treatment groups. This confirms that countries that sign TAs, regardless of their type, increase their bilateral trade.7

individual-level panel data estimation, considering that country pairs have signed TAs at different moments. We take as benchmark the gravity model (GM) to estimate bilateral trade flows and to address the effect of TAs on a balanced panel of 110 countries for the period 1995–2013. The GM has been widely used to explain bilateral trade flows using country size (GDP) and the geographical distance between two countries as the main explanatory variables. In addition, the GM allows the consideration of several other controls, such as trade barriers, cultural differences, and transaction and transportation costs. Gravity models have also been used to study the effect of IPRs on bilateral trade flows, finding mixed but significant effects (see, for example, Fink and Primo Braga, 2005; Campi and Dueñas, 2016), and to estimate the effect of trade agreements (see, for example, Carrere, 2006; Baier and Bergstrand, 2007). In our benchmark specification, wijt is the natural logarithm of exports from country i to country j at the year t, and the gravity equation is defined as:

wijt =

+ Xijt · + TAijtnip · + TAijtip · +

ij

+

t

+

ijt ;

(5)

where i, j = 1, …, N; α is a constant term; X = { ln(GDPi), ln(GDPj), hci, hcj, TRIPSi, TRIPSj} is a vector of country specific macro variables, including GDP – as an indicator of economic and market size – human capital – as an indicator of development or capabilities – and a variable indicating adherence to the TRIPS that takes the value of 1 since the year in which countries are in compliance with TRIPS and 0 otherwise8;

4.2. Gravity estimation using panel data

8 Countries that joined the WTO were given different transition periods for the implementation of IP laws and enforcement mechanisms. Developed countries were given one year (until 1996), developing countries and some transition economies were given five years, until 2000. Least developed countries initially had 11 years, until 2006. This transition period was extended to 2016 for pharmaceutical patents and undisclosed information, including trade secrets, and to 2013 for all other categories. In 2015, it was agreed to further extend exemptions on pharmaceutical patent and undisclosed information protection for least developed countries until 2033 or until when they cease to be a least developed country, whichever date is earlier. See: https://www.wto. org/english/theWTO_e/whatis_e/tif_e/agrm7_e.htm.

The second strategy to study the effect of TAs is to implement an 7 The spikes observed in the ATET estimations between 2004 and 2007 might be related with the enlargements of the European Union taking place in those years, with 10 and 2 countries becoming members, respectively. This generates changes in the control and treatment samples and results in an upward bias of the effect of TAs with IP chapters for the treated group. We estimated the ATETs excluding those countries from the sample and the spikes are notably reduced. The results are available upon request.

7

Research Policy xxx (xxxx) xxx–xxx

M. Campi, M. Dueñas

Table 2 Bilateral trade estimation results of the gravity model using panel data fixed effects (country-pairs) estimation method. Model TA nip

TA

(1)

(2)

(3)

(4)

(5)

(6)

0.069** (0.028)

0.134*** (0.045) 0.084*** (0.027)

0.131*** (0.044) 0.035 (0.027) 0.028

0.123*** (0.044) 0.026 (0.027) 0.028

(0.037) 0.131***

(0.036) 0.155***

(0.020)

(0.020) 0.121*** (0.018) 0.118*** (0.018) 1.125*** (0.053) 1.438*** (0.045) 0.069 (0.098) −0.131 (0.091) −21.993*** (0.874) Yes 201,769 0.209 11,919

***

0.098 (0.026)

TAip

0.164*** (0.053)

TAtnip5 TAtip 5 TRIPSi TRIPSj ln(GDPi) ln(GDPj) hci hcj Constant Time dummies Observations R-squared Number of links

1.129*** (0.053) 1.442*** (0.045) 0.185* (0.097) −0.016 (0.090) −22.505*** (0.859) Yes 201,769 0.205 11,919

1.145*** (0.057) 1.473*** (0.049) 0.172* (0.103) −0.044 (0.096) −23.245*** (0.950) Yes 180,178 0.194 11,397

1.145*** (0.056) 1.475*** (0.048) 0.223** (0.102) −0.030 (0.094) −23.275*** (0.926) Yes 188,720 0.200 11,467

1.127*** (0.053) 1.440*** (0.045) 0.184* (0.097) −0.017 (0.090) −22.452*** (0.862) Yes 201,769 0.205 11,919

1.138*** (0.053) 1.450*** (0.046) 0.176* (0.097) −0.026 (0.090) −22.676*** (0.866) Yes 201,769 0.206 11,919

Notes: The dependent variable is the ln of total bilateral trade. Clustered robust standard errors are in parenthesis. * Significance level: p < 0.10. ** Significance level: p < 0.05. *** Significance level: p < 0.01.

TAnip and TAip take the value of one since their signing and zero before; γij are country pairs fixed effects; τt are time dummies; and η is the residual. The model assumes that TA is strictly exogenous, and that it is not correlated with η. See Appendix C for a complete description of variables and sources. In order to deal with the endogeneity of TAs and to fix unobserved characteristics, we estimate the gravity equation applying a fixed effects estimation method on the panel data – this is controlling for bilateral economic and political relationship of country-pairs. This allows us to capture any particular tendency on the signing of TAs formed by the specific relations between two countries and geographical, cultural, political, or economic characteristics specific to the pair (Baier and Bergstrand, 2007; Cheong et al., 2015; Shin and Ahn, 2018). In addition, we use year dummies in order to capture macroeconomic trends in world trade flows. Another possible strategy is to use time varying countries fixed effects in order to consider the unobservable multilateral resistance terms. However, given that we allow for asymmetric effects on the country pairs and that we have a high number of countries and years, time varying countries fixed effects add a total of 4,180 dummies that absorb most of the effects of TAs in the current structure of the model. It is important to mention that these estimations are highly sensitive to the specification of the model and the groups of countries or years (Magee, 2003). Despite the use of country-time varying fixed effects can help addressing the issue of endogeneity, reducing the sample of countries or using a shorter period of time or reducing the number of links by considering that they are symmetric, can increase the problem of multicollinearity, which can be partly addressed by using a high number of observations because this increases the precision of the estimated coefficients. In brief, given this trade-off, we prefer to keep a large panel data and estimate it with country-pairs fixed effects and time dummies, including different robustness checks. As in the matching estimations, we use as the control group country

pairs that do not have any TA and two treatment groups of country pairs that have signed TAnip or TAip. All estimations present robust standard errors clustered by country pairs. Table 2 shows the estimation results of the gravity model for total bilateral trade. GDP of both the importer and the exporter are positive and significant. Human capital is positive and significant for the exporter, indicating that countries with higher capabilities are able to export more. In model (1) we include a variable indicating if country pairs have a TA regardless of the type. As expected, we estimate a positive and significant coefficient. In models (2) and (3) we include a variable indicating if the countries have signed a trade agreement with no IP chapters (TAnip) and with IP chapters (TAip), respectively. As in the matching estimations, we observe that signing any type of TA increases bilateral trade flows. In model (4), we include both variables in order to be able to compare their effect on bilateral trade flows. We find that both are positive and significant, but TAnip have a stronger effect. There are two plausible explanations for this result. In the first place, especially in recent years, TAnip are more frequently signed by pairs of countries with more similar characteristics, particularly by LDCs. Therefore, TAnip are more likely to arise endogenously between countries with higher latent trade propensities than do more expansive TAs, such as TAip, which often involve DC-LDC pairs. The second reason is that more comprehensive TAs, such as TAip often include a few trade issues compared to TAnip, while they focus to a greater extent on traderelated issues, including IPRs. In other words, it could be argued that TAnip include broader negotiations on trade issues that could create more opportunities for trade compared to TAip. Therefore, the effect of TAnip on bilateral trade flows is likely to be stronger than the effect of TAip. But also, other factors affecting the impact of TAs need to be considered. Trade agreements are phased-in over time, generally over five to ten years, as they include agreed upon phased-in tariff cuts. Baier and 8

Research Policy xxx (xxxx) xxx–xxx

M. Campi, M. Dueñas

Table 3 Estimation results of the gravity model for bilateral trade of high- and low-IP intensive products using panel data fixed effects (country-pairs) estimation method. High-IP intensive products Model nip

TA

TAip

(1)

Low-IP intensive products (2)

***

0.231 (0.049) 0.119*** (0.029)

TAtnip5 TAtip 5 TRIPSi

(3) ***

ln(GDPj) hci hcj Constant Country dummies Observations R-squared Number of links

1.138*** (0.052) 1.325*** (0.048) 0.212** (0.097) −0.005 (0.090) −22.388*** (0.875) Yes 185,203 0.266 11,754

(5) ***

0.130 (0.045) 0.006 (0.031) 0.069*

0.122*** (0.044) −0.002 (0.031) 0.069*

0.205 (0.047) 0.057** (0.028) 0.065*

(0.039) 0.134***

(0.039) 0.159***

(0.040) 0.139***

(0.040) 0.159***

(0.022)

(0.022) 0.203*** (0.018) 0.066*** (0.018) 1.107*** (0.053) 1.334*** (0.048) 0.026 (0.099) −0.078 (0.092) −21.062*** (0.873) Yes 185,203 0.269 11,754

(0.022)

(0.022) 0.103*** (0.019) 0.111*** (0.019) 1.043*** (0.058) 1.452*** (0.049) −0.006 (0.106) −0.129 (0.099) −20.606*** (0.947) Yes 194,657 0.138 11,863

1.147*** (0.052) 1.335*** (0.048) 0.202** (0.097) −0.017 (0.091) −22.585*** (0.879) Yes 185,203 0.266 11,754

0.148 (0.047) 0.059* (0.031)

(6) ***

0.213 (0.047) 0.069** (0.028) 0.068*

TRIPSj ln(GDPi)

(4) ***

1.045*** (0.057) 1.457*** (0.049) 0.098 (0.104) −0.019 (0.098) −22.121*** (0.944) Yes 194,657 0.135 11,863

1.055*** (0.057) 1.466*** (0.050) 0.087 (0.104) −0.030 (0.098) −22.318*** (0.949) Yes 194,657 0.135 11,863

Notes: The dependent variable is the ln of bilateral trade of high-IP intensive products (models 1–3) and bilateral trade of low-IP intensive products (models 4–6). Clustered robust standard errors are in parenthesis. * Significance level: p < 0.10. ** Significance level: p < 0.05. *** Significance level: p < 0.01.

Bergstrand (2007) showed that the effect of TAs can be seen after 10 years of the entry into force. In addition, trade-related issues, including IP provisions, might demand a relatively long implementation time, despite most of them are discussed and defined during the negotiation period. Thus, in order to consider a possible delay in the implementation time, in model (5), we include the five-year lag of the variables TAnip and TAip.9 Note that given that the data on TAs are available since 1948, we do not loose observations when we create the lags. We observe that TAtnip5 is not significant while TAtip 5 is positive and significant. This implies that, on average, the positive effect of TAip tends to be stronger five years after its entry into force. Usually, TAip are more comprehensive and have phased-in implementation times. It might take a longer time for this type of TAs to be completely implemented and trade issues depend on this implementation. Considering both the immediate and the lagged effects of TAs, model (5) estimates that a pair of countries signing a TAnip or a TAip will have a net increase of their bilateral trade flows of 17% and 18%, respectively.10 Finally, in model (6), we included a variable that controls the date in which countries have adhere to the TRIPS agreement. This variable aims to capture the effect of reforms related to TRIPS, while TAip is expected to capture the effect of TRIPS+. We observe that TRIPS is significant and positive for both the importer and the exporter, and that TAtip 5 does not lose its positive and significant effect. On the contrary, the net expected increase estimated is 16% from TAnip and 20% from TAip. Next, we estimate Eq. (5) for products that are expected to be

differently affected by IPRs. Table 3 shows the estimation results of the gravity model using trade flows of high-IP and low-IP intensive products as dependent variables. As before, GDP per capita of both the importer and the exporter are positive and significant. A higher human capital of the exporter increases exports of high-IP intensive products, which mainly include products from high-technology sectors, while it is not significant for trade of low-IP intensive products. In models (1) and (4), we observe that both TAnip and TAip increase bilateral trade but the effect of TAnip is again stronger than the effect of TAip, regardless of the IP intensity of traded products. If we consider the lagged effects of TAs, we observe in models (2) and (5) that the effect of TAtnip5 and TAtip 5 are significant for both types of products, but in both cases TAtip 5 has a stronger effect. The estimated net expected effect on bilateral trade derived from TAnip is stronger, 32% in high-IP intensive products and 22% in low-IP intensive products, against 23% and 17% respectively, derived from TAip. Models (3) and (6) include the variable indicating the date in which countries have complied with the requirements of the TRIPS. In both types of products, we observe that the effect of the TRIPS is positive and significant, and that including this variable does not change the significance and the positive effects of the variables related with TAip. Including the variable TRIPS does not reduce the estimated effect of the signing of TAs, and TAnip still has stronger expected effects (31% in high-IP and 21% in low-IP), compared to TAip (24% in high-IP and 17% in low-IP intensive products). Finally, we evaluate the effect of TAs on bilateral trade for country pairs of different development levels.11 We classify trade flows in four

9 We also estimated the model using 2, 3, and 10-year lags. The results provide similar conclusions and are available upon request. 10 The estimated average effect of a TAnip is exp(0.131 + 0.028) = 1.172 and of a TAip is exp(0.035 + 0.131) = 1.181, which correspond to increases of 17% and 18%, respectively. We have used the same methodology for the remaining changes in percentage presented in the paper.

11 The classification of countries according to their development level is based on United Nations (2017).

9

Research Policy xxx (xxxx) xxx–xxx

M. Campi, M. Dueñas

Table 4 Estimation results of the gravity model with interaction variables using panel data fixed effects (country-pairs) estimation method. Total bilateral trade Model

(1)

Gdc →dc · TA

nip

Gldc →dc · TAnip Gldc →ldc · TAnip Gdc →dc · TAip Gdc →ldc · TAip Gldc →dc · TAip Gldc →ldc · TAip

Gdc

nip dc ·TAt 5

Gdc

nip ldc ·TAt 5

Gldc

nip dc ·TAt 5

Gldc

nip ldc ·TAt 5

Gdc

ip dc ·TAt 5

Gdc

ip ldc ·TAt 5

Gldc

ip dc ·TAt 5

Gldc

ip ldc ·TAt 5

(2)

0.107 (0.072) 0.355*** (0.107) 0.118 (0.112) 0.113 (0.070) 0.199*** (0.043) 0.070 (0.045) −0.034 (0.056) 0.120 (0.081)

Gdc →ldc · TAnip

ln(GDPj) hci hcj Constant Time dummies Observations R-squared Number of links

(3) **

0.003 (0.071) 0.304*** (0.102) 0.041 (0.109) 0.161** (0.068) 0.125*** (0.043) 0.017 (0.045) −0.050 (0.058) 0.140* (0.073) 0.274***

0.218 (0.085) 0.223** (0.104) 0.507*** (0.125) 0.171** (0.079) 0.325*** (0.044) 0.042 (0.045) 0.126** (0.063) 0.064 (0.076)

Low-IP intensive products (4)

(5)

(6)

0.096 (0.087) 0.206** (0.094) 0.443*** (0.131) 0.194*** (0.074) 0.225*** (0.043) 0.009 (0.045) 0.094 (0.059) 0.084 (0.070) 0.318***

0.125 (0.083) 0.239** (0.094) −0.007 (0.122) 0.247*** (0.076) 0.131*** (0.049) −0.000 (0.056) −0.056 (0.062) 0.215** (0.092)

0.045 (0.079) 0.208** (0.089) −0.092 (0.119) 0.258*** (0.073) 0.077 (0.049) −0.073 (0.055) −0.062 (0.065) 0.204** (0.084) 0.214***

(0.049) 0.162*

(0.057) 0.064

(0.050) 0.104

(0.094) 0.239**

(0.090) 0.203*

(0.123) 0.263**

(0.107) −0.102*

(0.117) −0.043

(0.115) −0.014

(0.053) 0.143***

(0.058) 0.195***

(0.058) 0.107***

(0.023) 0.151***

(0.027) 0.102**

(0.026) 0.205***

(0.035) 0.050

(0.040) 0.096

(0.042) 0.024

(0.053) 0.003

(0.058) −0.012

(0.057) 0.039

(0.085) 1.099*** (0.048) 1.401*** (0.042) −0.035 (0.085) −0.254*** (0.081) −20.273*** (0.458) Yes 201,769 0.204 11,919

1.091*** (0.048) 1.396*** (0.042) −0.017 (0.085) −0.231*** (0.080) −20.218*** (0.456) Yes 201,769 0.204 11,919

ln(GDPi)

High-IP intensive products

1.251*** (0.047) 1.441*** (0.044) 0.210** (0.085) 0.010 (0.080) −25.186*** (0.463) Yes 185,203 0.264 11,754

(0.097) 1.257*** (0.047) 1.447*** (0.044) 0.186** (0.085) −0.015 (0.081) −25.212*** (0.464) Yes 185,203 0.264 11,754

0.930*** (0.051) 1.338*** (0.045) −0.235*** (0.091) −0.357*** (0.087) −17.272*** (0.492) Yes 194,657 0.133 11,863

(0.089) 0.937*** (0.051) 1.340*** (0.045) −0.250*** (0.091) −0.382*** (0.088) −17.294*** (0.495) Yes 194,657 0.133 11,863

Notes: The dependent variable is the ln of total bilateral trade (models 1–2), bilateral trade of high-IP intensive products (models 3–4), and bilateral trade of low-IP intensive products (models 5–6). Clustered robust standard errors are in parenthesis. * Significance level: p < 0.10. ** Significance level: p < 0.05. *** Significance level: p < 0.01.

indicate that TAnip increases the level of total bilateral trade from DCs to LDCs and TAip increases bilateral trade between DCs. Instead, we observe no significant effect for LDCs. If we now consider also the lagged effects of both TAnip and TAip, in model (2), we observe that the effect of signing any type of TA is always positive for DCs, as this increases trade flows between DCs and from DCs to LDCs. For flows of LDCs, TAtip 5 does not have a significant effect. Models (3) and (5) estimate the effect of TAs on trade flows of products of different IP-intensity. In the case of high-IP intensive products, we estimate that TAnip increases flows of all types of trade partners, while TAip increases flows between DCs, and from LDCs to DCs. In the case of low-IP intensive products, TAnip increases flows from DCs to LDCs and between LDCs. Instead, TAip increases flows between DCs and when both partners are LDCs.

groups: (i) trade between DCs; (ii) trade from DCs to LDCs; (iii) trade from LDCs to DCs; and (iv) trade between LDCs. We generalize Eq. (5) to derive interactions between TAs and these different groups. Thus,

wijt =

+ Xijt · + k

Gk · TAijtnip ·

k

+ k

Gk ·TAijtip ·

k

+

ij

+

t

+

ijt ;

(6)

where G is a binary variable indicating the trade group type k = {dc → dc, dc → ldc, ldc → dc, ldc → ldc}. Table 4 shows the estimation results of the gravity model that includes interaction variables between the level of development of trading partners and TAs for total bilateral trade flows (1–2), trade flows of high-IP intensive products (3–4), and trade flows of low-IP intensive products (5–6). For total bilateral trade, in model (1), the interaction variables 10

Research Policy xxx (xxxx) xxx–xxx

M. Campi, M. Dueñas

Fig. 5. Expected increase in bilateral trade flows from the signing of trade agreements with no IP chapters and with IP chapters for trade partners of different development levels.

Models (4) and (6) include interactions between the level of development of the country pairs and the 5-year lagged TAs. Some of the interaction variables in levels loose significance, although they do not change their signs, as the lagged variables gain significance. In the case of trade of high-IP intensive products, TAtnip5 has a positive effect on flows between DCs and from LDCs to DCs, while TAtip 5 has a positive effect on flows between DCs and from DCs to LDCs. In the case of trade of low-IP intensive products, we estimate positive effects of TAtnip5 for flows between DCs and from LDCs to DCs, and of TAtip 5 only for flows involving DCs. In none of the cases, trade flows from LDCs are significantly affected by TAtip 5 . Overall, from this analysis we can conclude that trade agreements increase bilateral trade but unevenly for developed and developing countries. The expected increases for country pairs of different development levels are heterogeneous. Fig. 5 shows the net expected gains for each group of countries and different types of products, considering both the immediate effect and the lagged effect of the signing of TAnip and TAip. Firstly, we observe a positive effect of trade agreements in most cases (except for TAip in the case of flows from LDCs to DCs in total bilateral trade and trade flows of low-IP intensity). For all trade flows between DCs and from DCs to LDCs, the effects are positive and tend to be stronger than the estimated increases for LDCs, in particular those derived from the signing of TAip. For trade flows between LDCs and from LDCs to DCs, TAnip has stronger effects. The effect of both types of TAs on trade flows between LDCs is low, except in low-IP intensive products. The highest expected increase is estimated for trade flows from LDCs to DCs, deriving from TAnip, although it is worth considering that these flows are not very relevant in terms of their share on total trade for those countries. In brief, we estimate a positive and significant effect of both types of TAs for bilateral trade flows between DCs and from DCs to LDCs, regardless of the IP intensity of products. Instead, from LDCs to DCs, we observe positive effects in the case of high-IP intensive products deriving from TAnip. From these estimated effects it is clear that for LDCs the gains derived from TAip are lower than the gains derived from TAnip.

these sectors because they are highly intensive in the use of IPRs and because IP chapters set down particularly rigorous standards for them. We use the same estimation strategy as before, this is, panel data fixed effects at the country pair level with time dummies, in order to control for endogeneity. Table 5 presents the estimations results. We observe interesting effects at this level of disaggregation. In the case of pharmaceutical products, we find a positive effect of TAtip 5 that is also present when including the variable TRIPS, in model (2). The coefficient of the variable TRIPS is negative, although small. This is not surprising considering that this sector was subject to several exceptions and longer transition periods, and a negative or not significant effect of TRIPS on pharmaceutical products was also found in other empirical studies (see, for example Delgado et al., 2013). Similarly, trade of chemical products increases with both TAtip 5 and TAtnip5 . In contrast to pharmaceutical products, the effect of TRIPS appears positive for chemicals. Finally, in the case of machinery and electrical equipment, including mostly high-technology products and, in particular, ICT, the positive effect of TAs is stronger and derives from both types of TAs. Additionally, the signing of the TRIPS is positive but only for the exporter. These robustness checks add to the idea that a stronger effect of TAip is expected for sectors that are more IP-intensive. Likewise, the analysis suggests that the effect of TRIPS varies by sector, as argued by Delgado et al. (2013), who also found that in sectors such as ICT, TRIPS stimulated trade, but in other sectors such as pharmaceuticals, the signing of the TRIPS did not have a great impact, or even it had a negative one. Next, we distinguish between TAip in which one partner is the US or the EU/EFTA, because these agreements usually include strong but also different types of IP provisions. As we already discussed, agreements in which one partner is the US emphasize on patent and copyright protection, and usually include requirements for test data confidentiality for pharmaceuticals and chemicals. Agreements in which one partner is the EU or EFTA tend to include more general and wider IP provisions, emphasize more on the protection of geographical indications, and more recently began including test data confidentiality for pharmaceuticals and chemicals like the US. In addition, most TAs of the EU include the ratification of WIPO treaties and, both TAs in which one partner is the US or the and EU, include adherence to UPOV 1991. Table 6 shows the estimation results, which are in line with previous estimations. Interestingly, in the variables considering TAs in which one partner is the US, we only observe a significant effect for pharmaceutical products and chemicals, which is consistent with the predominant content of those specific TAs. Instead, TAip in which one partner is the EU or EFTA have a positive effect in all the different types of products considered and, particularly, we observe a delay in the effect that is captured by the lags. As we argued, these types of agreements are wider in their IP-related content.

4.3. Further robustness checks Next, we conduct robustness checks to complement and add further evidence to our findings. Firstly, we analyze the effect of TAs on more disaggregated IP-intensive products. Secondly, we consider whether TAs with the US of the EU/EFTA have different effects on bilateral trade. We estimate the impact of different types of TAs for three sectors: pharmaceutical products – chapter 30 in the HS code – chemicals (excluding pharmaceutical products) – chapters 28 to 38, except 30, and machinery and electrical equipment – chapters 84 and 85 – that includes information and communication technologies (ICT). We choose

11

Research Policy xxx (xxxx) xxx–xxx

M. Campi, M. Dueñas

Table 5 Estimation results of the gravity model for bilateral trade of pharmaceutical products, chemicals, and machinery and electrical equipment using panel data fixed effects (country-pairs) estimation method. Pharmaceutical products

Chemical products

Machinery and electrical equipment

Model

(1)

(2)

(3)

(4)

(5)

(6)

TAnip

−0.019 (0.068) 0.000 (0.044) 0.003

−0.014 (0.068) 0.008 (0.044) 0.005

0.016 (0.054) 0.041 (0.037) 0.092*

0.008 (0.054) 0.028 (0.037) 0.088*

0.195*** (0.054) 0.091*** (0.032) 0.100**

0.182*** (0.054) 0.076** (0.032) 0.094**

(0.054) 0.234***

(0.055) 0.225***

(0.050) 0.179***

(0.050) 0.199***

(0.044) 0.171***

(0.043) 0.198***

(0.038)

(0.038) −0.078*** (0.029) −0.045 (0.028) 0.622*** (0.078) 0.288*** (0.071) 0.119 (0.165) 0.054 (0.144) −5.144*** (1.331) Yes 103,383 0.172 8,736

(0.029)

(0.029) 0.150*** (0.022) 0.087*** (0.021) 1.093*** (0.068) 1.033*** (0.056) 0.281** (0.122) −0.137 (0.107) −20.460*** (1.155) Yes 138,211 0.117 10,347

(0.027)

(0.027) 0.313*** (0.021) 0.024 (0.021) 1.177*** (0.056) 1.401*** (0.055) 0.303*** (0.113) −0.242** (0.106) −24.614*** (1.011) Yes 162,635 0.272 11,358

TAip

TAtnip5 TAtip 5 TRIPSi TRIPSj

0.574*** (0.077) 0.285*** (0.071) 0.041 (0.161) 0.020 (0.143) −4.255*** (1.311) Yes 103,383 0.172 8736

ln(GDPi) ln(GDPj) hci hcj Constant Time dummies Observations R-squared Number of links

1.157*** (0.066) 1.042*** (0.057) 0.413*** (0.121) −0.062 (0.106) −21.776*** (1.136) Yes 138,211 0.116 10,347

1.276*** (0.056) 1.388*** (0.055) 0.582*** (0.112) −0.218** (0.105) −26.239*** (1.006) Yes 162,635 0.269 11,358

Notes: The dependent variable is the ln of bilateral trade of pharmaceutical products (models 1–2), bilateral trade of chemicals (excluding pharmaceutical products) (models 3–4), and bilateral trade of machinery and electrical equipment (models 5–6). Clustered robust standard errors are in parenthesis. * Significance level: p < 0.10. ** Significance level: p < 0.05. *** Significance level: p < 0.01. Table 6 Estimation results of the gravity model for bilateral trade of different types of products and different types of trade agreements using panel data fixed effects (countrypairs) estimation method.

Model nip

High-IP intensive products

Low-IP intensive products

Pharmaceutical products

Chemical products

Machinery and electrical equipment

(1)

(2)

(3)

(4)

(5)

(6)

***

TA

ip TAUS ip TAEU/EFTA

TAtnip5 ip TAUSt

Total bilateral trade

5

ip TAEU/EFTAt

5

ln(GDPi) ln(GDPj) hci hcj Constant Time dummies

***

**

0.159 (0.052) −0.072

0.234 (0.056) −0.110

0.133 (0.053) −0.021

0.086 (0.075) 0.110

0.073 (0.062) −0.156

0.164** (0.064) 0.071

(0.102) 0.009

(0.113) 0.064*

(0.115) −0.051

(0.154) −0.116**

(0.102) 0.051

(0.170) 0.069

(0.037) 0.007

(0.038) 0.073*

(0.043) 0.051

(0.057) 0.015

(0.050) 0.116**

(0.043) 0.099**

(0.041) 0.059

(0.043) 0.046

(0.045) 0.120

(0.060) 0.344***

(0.056) 0.295*

(0.048) 0.110

(0.082) 0.118***

(0.109) 0.109***

(0.083) 0.134***

(0.117) 0.233***

(0.160) 0.126***

(0.178) 0.097***

(0.027) 1.148*** (0.055) 1.483*** (0.048) 0.188* (0.100) −0.017 (0.093) −23.411*** (0.921) Yes

(0.029) 1.143*** (0.055) 1.371*** (0.050) 0.197** (0.100) −0.003 (0.094) −23.179*** (0.938) Yes

(0.030) 1.066*** (0.060) 1.494*** (0.052) 0.094 (0.108) −0.030 (0.101) −22.943*** (1.014) Yes

(0.054) 0.623*** (0.081) 0.478*** (0.075) 0.062 (0.169) 0.060 (0.148) −7.727*** (1.420) Yes

(0.039) 1.190*** (0.071) 1.103*** (0.060) 0.427*** (0.126) −0.105 (0.110) −23.090*** (1.237) Yes

(0.036) 1.266*** (0.059) 1.430*** (0.058) 0.558*** (0.116) −0.205* (0.108) −26.795*** (1.078) Yes

(continued on next page)

12

Research Policy xxx (xxxx) xxx–xxx

M. Campi, M. Dueñas

Table 6 (continued) Total bilateral trade

High-IP intensive products

Low-IP intensive products

Pharmaceutical products

Chemical products

Machinery and electrical equipment

Model

(1)

(2)

(3)

(4)

(5)

(6)

Observations R-squared Number of links

191,358 0.199 11,600

174,843 0.257 11,432

184,270 0.130 11,543

93,996 0.155 8393

128,187 0.110 10,019

152,380 0.259 11,032

Notes: The dependent variable is the ln of bilateral trade of total trade (model 1), and different types of products (models 2–6). Clustered robust standard errors are in parenthesis. * Significance level: p < 0.10. ** Significance level: p < 0.05. *** Significance level: p < 0.01.

5. Concluding remarks

We have also controlled for the date in which countries have complied with the demands of the TRIPS agreement. The effect of TAs with IP chapters is still positive and significant, implying that they are reflecting the effect of the reforms related to TRIPS+. In addition, we found that the effects of TAs also depend on the IP-intensity – and technology content – of products. Likewise, as Delgado et al. (2013), we found that the effect of TRIPS varies by sector depending on the exceptions and transitions periods provided for different IP-intensive products. Also, we found that the effects are heterogeneous for signatory countries of different development levels. We found a clear positive effect for developed countries in all types of products and to all destinations derived from both types of TAs. Instead, the gains for developing countries are weaker and the higher expected increases derive from TAs with no IP chapters. For LDCs, TAs with IP chapters mostly increase trade flows of low-IP intensive products between LDCs. From this, we could argue that IP chapters in TAs might be reinforcing the role of IPRs as trade barriers for developing countries. Overall, different estimation strategies and several robustness checks allows us to conclude that trade agreements increase bilateral trade but unevenly for developed and developing countries. In particular, DCs benefit from both types of TAs, while LDCs can get lower gains from TAs with IP chapters, which are increasingly leading to stronger and harmonized systems of IPRs, neglecting that countries with different capabilities might need different types of IPRs systems in order to enhance innovation and economic growth (Kim et al., 2012). Moreover, given that IP provisions and their related reforms imply real challenges for LDCs, the results of this paper raise the question of whether trade gains can compensate the effort related with IP reforms for developing countries.

During the last twenty five years there has been an increase in the number of trade agreements containing IP chapters as well as in the number of signatory countries, which has contributed to the strengthening and harmonization of IPRs systems beyond the process triggered by the TRIPS. In this framework, IP-demanding countries are usually developed countries, while developing countries are the ones that need to implement the reforms. This implies for LDCs loosing flexibility in the design of their IPRs systems that could best fit their needs and also important challenges in the implementation process. Thus, the motivations for accepting these IP reforms might be guided by the expected gains derived from trade issues of trade agreements. Despite the increasing relevance of TAs with IP chapters and their possible implications, this issue has been marginally addressed in empirical analysis, except from the recent study of Maskus and Ridley (2016) that analyzed the effect of TAs with IP chapters on total trade. In this paper we presented one of the first explorations of how trade agreements with IP chapters affect bilateral trade flows, considering products of different IP-intensity and countries of different development level. We used matching econometrics to compare the effect of TAs and TAs with IP chapters on signatory countries with respect to a control group of countries that did not sign TAs. The matching estimations showed that country pairs with any type of TAs have on average higher levels of trade than country pairs with no TAs. But we found no statistically significant differences on how different types of agreements affect trade. The estimations with panel data that capture fixed effects provide a better framework to assess and evaluate the differences between the effects of each type of agreements. Also, they allow for the consideration of possible asymmetric effects for countries of different development levels. We estimated a gravity model with fixed effects and a difference-in-difference technique, and we found that both types of TAs increase bilateral trade but TAs with no IP provisions have a stronger immediate effect while TAs with IP chapters have stronger effects after five years of their signing, which can imply that these TAs need a longer implementation time. Overall, the net expected increase of each type of TA is quite similar.

Acknowledgments We thank the editor and two anonymous reviewers for their suggestions that have helped us to improve our study. Earlier versions of this paper were presented at EMAEE 2017 (Strasbourg, France), MEIDE 2017 (Montevideo, Uruguay), and RIDGE Forum 2017 (Buenos Aires, Argentina). We thank the participants at these conferences for providing useful comments.

Appendix A. List of countries Developed countries Australia (AUS), Austria (AUT), Bulgaria (BGR), Canada (CAN), Croatia (HRV), Cyprus (CYP), Czech Rep. (CZE), Denmark (DNK), Estonia (EST), Finland (FIN), France (FRA), Germany (DEU), Greece (GRC), Hungary (HUN), Iceland (ISL), Ireland (IRL), Israel (ISR), Italy (ITA), Japan (JPN), Latvia (LVA), Lithuania (LTU), Malta (MLT), Netherlands (NLD), New Zealand (NZL), Norway (NOR), Poland (POL), Portugal (PRT), Slovakia (SVK), Slovenia (SVN), Spain (ESP), Sweden (SWE), Switzerland (CHE), United Kingdom (GBR), United States (USA). Developing countries Albania (ALB), Algeria (DZA), Angola (AGO), Armenia (ARM), Argentina (ARG), Azerbaijan (AZE)*, Bahamas (BHS)*, Bahrain (BHR), Bangladesh (BGD), Belarus (BLR)*, Bolivia (BOL), Bosnia Herzegovina (BIH)*, Brazil (BRA), Brunei Darussalam (BRN), Cambodia (KHM), Cameroon 13

Research Policy xxx (xxxx) xxx–xxx

M. Campi, M. Dueñas

(CMR), Chile (CHL), China (CHN), Colombia (COL), Congo (COG), Costa Rica (CRI), Ivory Coast (CIV), Dominican Rep. (DOM), Ecuador (ECU), Egypt (EGY), El Salvador (SLV), Equatorial Guinea (GNQ)*, Ethiopia (ETH), Gabon (GAB), Georgia (GEO)*, Ghana (GHA), Guatemala (GTM), Guinea (GIN)*, Honduras (HND), Hong Kong (HKG), India (IND), Indonesia (IDN), Iran (IRN), Iraq (IRQ), Jamaica (JAM), Jordan (JOR), Kazakhstan (KAZ), Kenya (KEN), Rep. of Korea (KOR), Kuwait (KWT), Kyrgyzstan (KGZ), Lebanon (LBN)*, Macao (MAC), Madagascar (MDG), Malaysia (MYS), Mauritius (MUS), Mexico (MEX), Rep. of Moldova (MDA), Morocco (MAR), Mozambique (MOZ), Myanmar (MMR), Nicaragua (NIC), Nigeria (NGA), Oman (OMN)*, Pakistan (PAK), Panama (PAN), Paraguay (PRY), Peru (PER), Philippines (PHL), Qatar (QAT), Russian Federation (RUS), Saudi Arabia (SAU), Senegal (SEN), Singapore (SGP), South Africa (ZAF), Sri Lanka (LKA), Syria (SYR), TFYR of Macedonia (MKD)*, Thailand (THA), Trinidad and Tobago (TTO), Tunisia (TUN), Turkey (TUR), Turkmenistan (TKM)*, Ukraine (UKR), United Arab Emirates (ARE), United Rep. of Tanzania (TZA), Uruguay (URY), Uzbekistan (UZB)*, Venezuela (VEN), Viet Nam (VNM), Yemen (YEM), Zambia (ZMB), Zimbabwe (ZWE). Note: * denotes countries which are included in the network analysis but are excluded from the econometric estimations because of the lack of data. Appendix B. Classification of exports according to IP intensity High-patent products (most of which are also high-trademark) Crude fertilizers Organic & Inorganic chemicals Dyeing materials Medicinal & pharmaceutical products Essential oils & perfume materials Chemical materials & products Rubber manufactures Power-generating machinery Machinery for industries

Metalworking machinery General machinery Office machines Telecommunications Electrical machinery Professional apparatus Photographic apparatus Miscellaneous mfg.

High-trademark products (with low-patent/copyrights) Dairy products & beverages Crude rubber Road vehicles Furniture Footwear

Manufactures of metals Pulp & waste paper Plastics Paper & related articles

High-copyright products (most of which are also high-trademark) Cinematographic film

Printed matter & recorded media

Appendix C. Description and sources of the variables employed in the econometric estimations Label

Related to

Description

Link

Exports (in ln) in constant (2000) US dollars

Source

BACI-CEPII: Gaulier and Zignago (2010) TAnip Link Trade agreements with no IP chapters Kohl et al. (2016) TAip Link Trade agreements with legally enforceable IP chapters Kohl et al. (2016) GDP Country Gross domestic product Penn World Tables: Feenstra et al. (2013) d Link Distance between two countries, based on bilateral distances between their largest BACI-CEPII: Gaulier and Zignago cities, weighted by the share of the city in the overall country's population (2010) contig Link Contiguity dummy equal to 1 if two countries share a common border BACI-CEPII: Gaulier and Zignago (2010) comlang Link Dummy equal to 1 if both countries share a common official language BACI-CEPII: Gaulier and Zignago (2010) Gk Link Set of dummies characterizing the development level of trading partners United Nations (2017) hc Country Index of human capital that considers the average years of schooling and the returns to Feenstra et al. (2013) education TRIPS Country Dummy variable that takes the value of 1 since the year in which countries are in Delgado et al. (2013), Park (2008), compliance with TRIPS and 0 otherwise Maskus and Ridley (2016)

w

14

Research Policy xxx (xxxx) xxx–xxx

M. Campi, M. Dueñas

References

Helpman, E., 1993. Innovation, imitation, and intellectual property rights. Econometrica 61 (6), 1247–1280. Hofmann, C., Osnago, A., Ruta, M., 2018. The content of preferential trade agreements. World Trade Rev. 1–34. Horn, H., Mavroidis, P.C., Sapir, A., 2010. Beyond the WTO? An anatomy of EU and US preferential trade agreements. World Econ. 33 (11), 1565–1588. Kim, Y.K., Lee, K., Park, W.G., Choo, K., 2012. Appropriate intellectual property protection and economic growth in countries at different levels of development. Res. Policy 41 (2), 358–375. Kohl, T., Brakman, S., Garretsen, H., 2016. Do trade agreements stimulate international trade differently? Evidence from 296 trade agreements. World Econ. 39 (1), 97–131. Kohl, T., Trojanowska, S., 2015. Heterogeneous trade agreements, WTO membership and international trade: an analysis using matching econometrics. Appl. Econ. 47 (33), 3499–3509. Krugman, P., 1993. Regionalism versus multilateralism: analytical notes. In: Melo, J.D., Panagariya, A. (Eds.), New Dimensions in Regional Integration. Cambridge University Press, Cambridge. Liu, M., La Croix, S., 2015. A cross-country index of intellectual property rights in pharmaceutical inventions. Res. Policy 44 (1), 206–216. Magee, C., 2003. Endogenous preferential trade agreements: an empirical analysis. Contrib. Econ. Anal. Policy 2 (1). Maskus, K.E., 2015. Intellectual property in a globalizing world: issues for economic research. Asia-Pac. J. Account. Econ. 22 (3), 231–250. Maskus, K.E., Penubarti, M., 1995. How trade-related are intellectual property rights? J. Int. Econ. 39 (3), 227–248. Maskus, K.E., Ridley, W., 2016. Intellectual Property-Related Preferential Trade Agreements and the Composition of Trade. Robert Schuman Centre for Advanced Studies Research Paper No. 2016/35. Available at SSRN: https://ssrn.com/abstract= 2870572. Mattoo, A., Mulabdic, A., Ruta, M., 2017. Trade Creation and Trade Diversion in Deep Agreements. World Bank Policy Research Working Paper No. 8206. Medvedev, D., 2012. Beyond trade: the impact of preferential trade agreements on FDI inflows. World Dev. 40 (1), 49–61. Mercurio, B., 2006. TRIPS-plus provisions in FTAs: recent trends. In: Lorand Bartels, F. (Ed.), Regional Trade Agreements and the WTO Legal System. Oxford University Press, Oxford. Park, W., 2008. International patent protection: 1960–2005. Res. Policy 37 (4), 761–766. Rose, A.K., 2004. Do we really know that the WTO increases trade? Am. Econ. Rev. 94 (1), 98–114. Schneider, P.H., 2005. International trade, economic growth and intellectual property rights: a panel data study of developed and developing countries. J. Dev. Econ. 78 (2), 529–547. Serrano, A., Boguñá, M., 2003. Topology of the world trade web. Phys. Rev. E 68, 015101(R). Shin, W., Ahn, D., 2018. Trade gains from legal rulings in the WTO dispute settlement system. World Trade Rev. 1–31. Shin, W., Lee, K., Park, W.G., 2016. When an importer's protection of IPR interacts with an exporter's level of technology: comparing the impacts on the exports of the North and South. World Econ. 39 (6), 772–802. Subramanian, A., Wei, S.-J., 2007. The WTO promotes trade, strongly but unevenly. J. Int. Econ. 72 (1), 151–175. United Nations, 2017. Development Status Groups and Composition. Available at: http:// unctadstat.unctad.org/EN/Classifications/DimCountries.

Abadie, A., Imbens, G.W., 2006. Large sample properties of matching estimators for average treatment effects. Econometrica 74 (1), 235–267. Almog, A., Bird, R., Garlaschelli, D., 2017. Enhanced Gravity Model of Trade: Reconciling Macroeconomic and Network Models. Available at: http://arxiv.org/abs/1506. 00348. Anderson, J., 1979. A theoretical foundation for the gravity equation. Am. Econ. Rev. 69 (1), 106–116. Arpino, B., Benedictis, L.D., Mattei, A., 2017. Implementing propensity score matching with network data: the effect of the General Agreement on Tariffs and Trade on bilateral trade. J. R. Stat. Soc. Ser. C (Appl. Stat.) 66 (3), 537–554. Baier, S.L., Bergstrand, J.H., 2004. Economic determinants of free trade agreements. J. Int. Econ. 64 (1), 29–63. Baier, S.L., Bergstrand, J.H., 2007. Do free trade agreements actually increase members’ international trade? J. Int. Econ. 71 (1), 72–95. Baier, S.L., Bergstrand, J.H., 2009. Estimating the effects of free trade agreements on international trade flows using matching econometrics. J. Int. Econ. 77, 63–76. Biadgleng, E.T., Maur, J.-C., 2011. The Influence of Preferential Trade Agreements on the Implementation of Intellectual Property Rights in Developing Countries: A First Look. UNCTAD-ICTSD Project on IPRs and Sustainable Development Paper No. 33. Available at SSRN: https://ssrn.com/abstract=1962832. Cameron, A.C., Trivedi, P.K., 2005. Microeconometrics. Cambridge University Press, Cambridge. Campi, M., Dueñas, M., 2016. Intellectual property rights and international trade of agricultural products. World Dev. 80, 1–18. Campi, M., Nuvolari, A., 2015. Intellectual property protection in plant varieties. A new worldwide index (1961–2011). Res. Policy 4 (44), 951–964. Carrere, C., 2006. Revisiting the effects of regional trade agreements on trade flows with proper specification of the gravity model. Eur. Econ. Rev. 50 (2), 223–247. Cheong, J., Kwak, D.W., Tang, K.K., 2015. Heterogeneous effects of preferential trade agreements: how does partner similarity matter? World Dev. 66, 222–236. Delgado, M., Kyle, M., McGahan, A.M., 2013. Intellectual property protection and the geography of trade. J. Ind. Econ. 61 (3), 733–762. Dueñas, M., Fagiolo, G., 2013. Modeling the international-trade network: a gravity approach. J. Econ. Interact. Coord. 8 (1), 155–178. Dür, A., Baccini, L., Elsig, M., 2014. The design of international trade agreements: introducing a new dataset. Rev. Int. Organ. 9 (3), 353–375. Falvey, R., Foster-McGregor, N., 2017. On the Relationship between the Breadth of PTAs and Trade Flows. United Nations University-Maastricht Economic and Social Research Institute on Innovation and Technology (MERIT). Working Paper Series. 2017-038. Feenstra, R.C., Inklaar, R., Timmer, M.P., 2013. The Next Generation of the Penn World Table. Available for download at: www.ggdc.net/pwt. Fink, C., Primo Braga, C.A., 2005. How stronger protection of intellectual property rights affects international trade flows. In: Fink, C., Maskus, K.E. (Eds.), Intellectual Property and Development: Lessons from Recent Economic Research. World Bank Publications. Gaulier, G., Zignago, S., 2010. BACI: International Trade Database at the Product-Level. The 1994–2007 Version, Working Paper. CEPII Research Center. Available at: http:// ideas.repec.org/p/cii/cepidt/2010-23.html. Grossman, G.M., Helpman, E., 1990. Trade, innovation, and growth. Am. Econ. Rev. 80 (2), 86–91. Grossman, G.M., Lai, E.L.-C., 2004. International protection of intellectual property. Am. Econ. Rev. 94 (5), 1635–1653.

15